# haarnoja/sac

Soft Actor-Critic

Repository: https://github.com/haarnoja/sac
Canonical: https://ross.abutalabs.com/products/sac
Language: Python
License: NOASSERTION
License Family: other
Last push: 2023-11-29T20:49:46+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 3192, "days_push": 1008, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1301, forks 252 (observed 2026-08-28T04:04:17.662054+00:00)

## What it is
The original reference implementation of Soft Actor-Critic (SAC), a deep reinforcement learning algorithm for training maximum entropy policies in continuous action domains, built on TensorFlow. The repository is no longer maintained and points users to the successor softlearning package.

## Use cases
- train maximum entropy policies for continuous control tasks
- reproduce results from the SAC ICML 2018 paper
- run reinforcement learning experiments in Mujoco environments
- learn diverse skills with DIAYN
- study an off-policy actor-critic algorithm implementation
- benchmark continuous-domain RL agents

## When to choose
- you need the exact original TensorFlow SAC implementation for research reproduction
- you want to extend or study the reference code behind the SAC paper
- you need DIAYN-based diverse skill learning on top of SAC

## When to avoid
- you want an actively maintained SAC implementation - use softlearning or a PyTorch alternative like rlkit
- you prefer PyTorch over TensorFlow
- you need modern dependency versions or recent bug fixes
- you want to avoid the legacy rllab and Mujoco setup burden

## Facets
- artifact type: library
- maturity: abandoned
- function: reinforcement-learning, machine-learning, deep-learning
- domain: reinforcement-learning, machine-learning, robotics
- platform: python
- tags: soft-actor-critic, maximum-entropy-policy, tensorflow, mujoco, continuous-control, off-policy-learning, diayn, superseded-by-softlearning, docker, linux, macos

## Member repositories
- haarnoja/sac (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:17.662054+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T04:52:15.001907+00:00, confidence not recorded.
  - readme: https://github.com/haarnoja/sac (fetched 2026-08-28T04:04:17.662054+00:00, sha 2c0b76578a94)
- Data as of 2026-08-30T08:39:29.467469+00:00.